I recently spoke with Palo Alto Networks CIO Meerah Rajavel about how the company is using AI to drive both internal innovation and defend against a new generation of AI threats.
As a founding investor and former board member of Palo Alto Networks, I’ve had a front-row seat to its evolution, and this conversation offered a timely look at how it’s helping protect enterprises in the AI era.
This is the eighth installment in a series of interviews with leading CIOs across global enterprises.
Asheem Chandna: How has Palo Alto Networks approached AI and what impact do you see on the security landscape?
Meerah Rajavel: We are in the midst of a massive AI transformation. At Palo Alto Networks, we believe that innovation – with AI at its core – is our most powerful weapon to stay a step ahead of our adversaries. This isn’t just a buzzword; it’s about fundamental shifts in how we protect ourselves.
We approach AI through two lenses. First, as a cybersecurity provider, we focus on enabling secure use of AI for our customers to drive their superior business outcomes.
Second, we’re combatting AI with AI. As adversaries adopt these tools, we’re leveraging them ourselves to stay ahead of evolving threats. Machine learning powers high-volume tasks like pattern matching. Deep learning drives predictive capabilities, which is critical in security. Gen AI is transforming user interaction, making tools more intuitive and responsive.
We apply this strategy across our products and internal operations. For example, our developers have been running agentic-style AI workloads for the past nine months. AI is both integrated in what we deliver and how we work internally.
How has Gen AI changed the adversary and threat landscape? How has that shaped your response?
There is an ongoing race between attackers and defenders. Hackers leverage AI to create faster, more sophisticated attacks. For example, Gen AI has dramatically lowered the barrier to attack. Take phishing campaigns, which have become far more convincing and now span beyond email into voice, video, and deepfakes.
The speed of attacks has also surged. Our Unit 42 research team found that in a quarter of cases, the point of compromise to data loss was less than 5 hours, three times faster than in 2021. And in 19% of cases it was less than one hour.
We can only tackle these challenges using AI. And we are making progress. Dwell time—the window between breach and detection—has dropped to near real-time.
At the same time, the tech stack itself is evolving: new hardware, vector databases, unstructured data, and shifting access controls create greater complexity. Many open-source models are jailbroken, and prompt injection is an increasing concern. To help secure our customers’ AI business transformations, we’ve launched Prisma® AIRS™, the industry’s most comprehensive AI security platform. It is designed to protect a company’s entire AI ecosystem: AI apps, agents, models, and data.
We also continue to strengthen our AI-powered network and real-time security operations platforms to help companies stay ahead of attackers in the age of AI.
Is AI creating entirely new categories of threats?
It’s often less about entirely new threats and more about mutation—combinations of existing techniques that make detection harder. That’s driven a spike in zero-days. Among our 70,000 customers, we detect about 11.3 billion threats daily. A few months ago, 2.5 million of those were zero-days.
We’re also seeing attacks on AI infrastructure itself. Take the Air Canada incident last year, not a cyberattack, but a flawed prompt caused a real-world failure. Used maliciously, that same vector could trigger brand damage or financial loss.
Emerging categories include model manipulation, prompt injection, and infrastructure exploits. This is still in the early-stages, but growing fast.
AI itself is our greatest hope for getting ahead of these new threats. It allows for predictive threat intelligence, identifying attacks before they launch, and automating response at machine speed—something human teams can never do.
Where has AI created the most internal value beyond security?
We saw early that this wasn’t just another tech wave—it was a tectonic shift. In late 2022, we began hosting monthly sessions with 50+ senior leaders to demystify AI. We brought in experts from Google, VC firms, and others at the forefront of their field.
That led to an internal “AI Mastermind Challenge” with real prizes. Employees could submit ideas in any area, not just security. We received over 850 submissions, with standout entries from product, finance, marketing, IT, and sales.
That grassroots creativity helped shape our product roadmap and internal tools. From the top down, we also identified four priority areas for AI and automation. One was enterprise ticketing—across IT, HR, and finance—which handles over 480,000 tickets a year.
Our CEO, Nikesh Arora, set an ambitious goal: eliminate 80-90% of these tickets with AI.
At first, we thought Gen AI alone could handle it. But only 18% of tickets involved summarization or retrieval. The breakthrough came from combining cutting-edge Gen AI tools with traditional automation—what I jokingly call our “poor man’s agentic AI.” That became Panda AI, which we launched in mid-2023. It now handles all IT, HR, and finance questions and requests—no phone lines, no portals, and minimal Slack channels. Just one seamless, AI-powered experience.
Essentially, we are now using AI to solve IT problems employees face that have been solved many times before. Now, we’ve automated the response and freed up those resources–whether contractors or employees–to focus on more complex issues and tasks.
Can you share a workflow that’s been transformed by AI?
Take IT software access. It used to be a manual process: submit a request, trigger a workflow, get approvals, and check RBAC (role-based access control).
The traditional approach lacked nuance. To address that, we built an ML model that analyzed org structure, HR data, and usage patterns to create smarter access rules.
Now, low-risk access is fully automated. Higher-risk access may be routed for review—but even that’s faster. We use our own product, Cortex XSOAR, to drive all automation behind the scenes.
Looking ahead three years, how do you see AI evolving internally?
We are thinking about AI not just as a technology, but about how AI will shape both existing and new processes; in particular, people processes are critical and need to be considered. This will also lead to new roles, such as in AI engineering and governance.
Today, we are using AI along with Palo Alto Networks’ products to drive internal efficiencies and evolution. We’re starting in engineering, but the broader goal is breaking down enterprise workflows into atomic components that AI can automate.
Gen AI already supports coding and unit testing. Over 6,000 engineers use our internal models, hosted in-house to protect IP. Even R&D teams are firewalled from each other’s data.
We’ve also rolled out secure GPT models to help employees reduce mundane tasks and experiment with productivity hacks. Moving through this change, we’ve been thoughtful about how we celebrate our company values of disruption, integrity, and execution through the enablement process. We see employees gaining more opportunities to shift their focus from the task-level, giving them space and time to think about the biggest challenges we are solving and building on those even faster.
The next step is full lifecycle automation—going from ideation to deployment with minimal human touch. Software development is structured, which makes it ideal.
Agentic AI is the current buzzword, but in our case, it’s about speed and efficiency. I envision specialized agents: a PM agent to refine user stories, a QA agent, an Engineer agent, and more.
Right now, we’re focused on the Software Development Lifecycle—and this could reshape how software is built across the entire enterprise.
Where has AI had the biggest workforce impact, positive or negative?
We embedded AI into how we resolve every single technical support case for our customers. We focused first on technical support, where we saw a massive opportunity to delight our customers and respond to cases much more quickly. These cases are technically complex and high stakes. We knew that if we could use AI to help our teams handle the toughest cases, we would unlock real transformation. This required us to wholly understand how every single case got resolved and use AI to essentially learn from every interaction. By focusing on the full system and understanding in detail the data and expertise used to resolve cases, it helped us use AI to have a bigger impact from the start.
Internally, the biggest disruption so far has been among contractors, especially in IT and support roles. We’ve reduced reliance there while reskilling internal teams for higher-value work.
For example, our service desk team once handled live support. Now they maintain the documentation that powers our AI tools; these are new roles that didn’t exist before. We’ve also built continuous learning loops: humans review and improve AI-generated outputs.
The next wave of disruption is coming—especially as we apply AI to HR and finance systems. We have to address privacy and data sharing concerns, which are valid.
Our employee experiences can also be enhanced with the use of AI. We’ve always offered an accelerated career experience with tons of learning. We see the opportunity to clear the way for more collaboration, more building together, and less administrative overhead as a huge win.
Until evaluation frameworks mature, humans need to stay in the loop. But those frameworks are improving fast. The bigger shift will be reskilling the workforce to keep pace.
What’s your experience working with Gen AI startups?
Mixed. I love working with startups; they often spot things others miss. But Gen AI is moving so fast that many can’t plant their feet. By the time many startups go to market, the landscape’s changed. I’ve asked about a value prop and gotten three different answers in a month. That makes it hard to build long-term trust.
What we really need are startups building the plumbing: evaluation frameworks, feedback loops, learning infrastructure, and governance systems. Not flashy—but essential pieces of the puzzle. And we need stability, even as architectures shift from LLMs to agents to reasoning engines. These are areas where we’d love to partner with startups. But no one’s nailed them yet, so we’ve had to build a lot of tools in-house.
We expect to see more and more partnerships emerge between startups and larger companies like Palo Alto Networks to accelerate AI advancements. Big companies can help startups access critical data in exchange for fresh ideas and agile innovations.
Startups are dealing with the same velocity we are. And to be honest, as customers, we’re part of the problem—we’re always asking, “What’s next?”
The winners will be the ones who go deep, stay focused, and build for the long haul.
What advice do you have for companies just beginning their AI journey?
Start simple. Pick a repeatable, workflow-based use case that improves user experience. Don’t lead with “AI”. Show the value: faster, easier, better.
That’s what made Panda AI successful. We didn’t announce it. We just turned it on and people noticed the improvement right away.
Same with sales. We built a Gen AI-powered Slack agent that helps 4,000+ go-to-market team members navigate product, pricing, and POC details. It replaced 120 Slack channels full of “Does anyone know…?” questions. Now there’s one agent, one channel, and a group of SMEs. When someone corrects the agent, that feedback is ingested within 15 minutes. It’s a live, evolving system.
Don’t overlook security as you embark on your AI journey. You can embrace AI with confidence if you know that what you are building is secure.
So again: focus on outcomes, not just tech. And show ROI. AI isn’t cheap—if you’re a CIO investing in AI, you need to prove value early and often.
